Swaged Needles: Inspection Strategy For A Sealed Joint
Nov 16, 2026
Pain Points: Everything Important Is Inside the Part
A swaged tip is a joint that nobody can see. The bore is closed by the forming operation, the wall is thickened or thinned depending on bore support, residual stress is locked into the transition, and the surface that will contact the patient is the one surface that cleaning cannot easily reach. Every property that governs performance is either internal, sub-millimetre or both, which makes conventional incoming inspection close to useless and turns quality into a matter of process confidence rather than product proof.
The consequence is a chronic mismatch between what is inspected and what fails. Dimensional reports pass, cosmetic inspections pass, and complaints still arrive about particulate, about bends that break at the shoulder, about flow rates below specification. Each of those is a different internal property: cleanliness state, residual stress, flow area. None of them is visible on the inspection sheet, because the inspection sheet was written for a machined part rather than a formed one.
The economic argument for fixing this is straightforward. Inspection is a cost centre that produces no value; destructive testing on samples produces information but destroys parts. A strategy that spends the inspection budget on the properties that actually fail, and uses sample-based destructive methods where the property cannot be measured non-destructively, is both cheaper and more defensible in an audit.
Working Principle: What Can Be Proved, and What Can Only Be Inferred
Inspection of a formed part divides into three tiers by what the evidence can support. Direct measurement covers external geometry and outer diameter at stations, and is available on every part if the cycle time allows. Profile projection of the taper belongs here, and so does the position of the tip relative to the shoulder.
Indirect or non-destructive inference covers wall thickness through the part, using ultrasound or radiographic methods calibrated against a sectioned reference. It does not prove the wall everywhere; it proves it at the spot size of the probe. Treating an inferred value as a direct measurement is the most common overstatement in component inspection reports.
Sample-based destructive evidence covers the properties that cannot be seen: wall at every station, grain structure, folds and micro-cracks at the transition, concentricity of the bore relative to the cone. These are qualification methods, and their proper use is to establish and periodically re-confirm that the process is in the state that the non-destructive checks assume.
The strategy follows from those tiers. Non-destructive checks on every lot, destructive checks on a defined frequency and after every change, and traceability that links the two so that a lot certificate states which evidence tier supports which claim.
Classification: Four Inspection Families
Geometric inspection covers the outer profile: taper angle, taper length, tip diameter, shoulder position, blend radius. It is the cheapest and most repeatable family, and it is also the least informative about function.
Structural inspection covers wall thickness at stations, concentricity and the integrity of the transition. It requires sectioning or a calibrated through-wall method, and it is the family that predicts bend failure and splitting.
Functional inspection covers insertion force, flow rate and, where relevant, leak integrity. It is performed on samples with fixtures that reproduce the intended use, and it is the family that speaks the same language as the device designer.
Cleanliness inspection covers particulate count, residue and bioburden where applicable. It is the family most often omitted at component level and most often implicated in device-level complaints, particularly where lubricant residue survived forming inside a closed taper.
Practical Operation: A Budgeted Inspection Plan
Set the plan by failure mode, not by drawing dimension. List the ways this part can fail in the device, then assign each failure mode an inspection method, a frequency and an acceptance criterion. A wall thinning failure gets wall sampling at mid-taper; a bend failure gets a bend test; a particulate failure gets a cleanliness check on a defined extraction.
Keep the destructive frequency tied to change, not only to time. A new tube lot, a replaced mandrel, a re-dressed die or a new cleaning bath all warrant a sectioned reference. Time-based sampling alone will miss the change that matters, because the change is what moves the process.
Hold the reference set. Keep a sectioned, photographed reference from the approved state, and compare new sections against it at the same magnification. Photographs of a polished section at fixed illumination detect fold formation and wall drift faster than a numeric comparison, because they show the shape of the defect as well as its size.
Write the report so it can be audited. Each claim should name the method, the location, the sample size and the date, and each inferred value should name the reference used to calibrate it. Under ISO 13485 this is what distinguishes a validated inspection from a plausible one.
Justify the sample size instead of defaulting to it. The right number of parts to section is the number that would detect the defect you are actually worried about at the rate it is actually occurring, and that can be estimated from the history of the process. A shop that has never found a fold in fifty sections is not proving absence; it is proving that its sampling strategy is not sensitive enough to find one. Stating the assumption openly is more defensible than repeating an arbitrary number.
Field Experience: Findings That Repeat
The most frequent internal defect found on sectioning is a fold at the transition, a smeared surface layer rolled over itself rather than healed. It survives dimensional inspection, opens during aggressive cleaning, and produces both particulate and a crack origin. A light stress relief or a controlled electropolish after forming removes it, but only if the inspection plan looks for it.
The second is wall drift across lots without any dimensional change: the outer profile is identical, the wall is thinner, and the flow rate rises quietly until a functional test catches it. This pattern almost always follows a mandrel change, and tracking mandrel life by part count is the countermeasure that works.
The third is lubricant residue trapped inside the closed taper. It is invisible, survives rinsing, and shows up as a bioburden or particulate result at device level. Solvent extraction on a sample, with a defined limit, is the only reliable way to detect it before it becomes someone else's non-conformance.
A fourth pattern is worth recording because it changes how inspection is scheduled: the defects that matter rarely appear at the first article. A new die set produces clean sections for the first few hundred parts and then begins to show a shoulder fold as the cavity wears. Inspecting only at setup therefore measures the tool at its best. Sampling at a defined midpoint in the tool life, not only at the start, is the practical way to characterise a process that degrades gradually.
Summary
A swaged needle conceals its own quality, so inspection must be organised by failure mode and by evidence tier rather than by drawing dimension. Geometric, structural, functional and cleanliness families each answer a different question. Non-destructive checks on every lot, destructive references after every change, and a report that names method and location together produce a defensible inspection strategy for a joint that cannot be seen.
Outlook
Computed tomography is moving from research into routine first-article work for fine cannulae, because it resolves wall, bore and concentricity in one scan without sectioning. Coupled with automated functional fixtures and defined extraction methods for cleanliness, the inspection plan of the next few years will look more like a validated measurement system than a checklist. For suppliers, the competitive advantage will lie in being able to state precisely what each certificate proves.







